云边端协同下的多云机器人卸载策略研究

Research on Multi-Cloud Robot Offloading Strategy under Cloud-Edge-End Collaboration

如何充分利用中心云、边缘云和本地设备的计算资源,使得系统总代价尽可能降低,是云机器人任务卸载领域亟须解决的主要问题之一。针对该问题,提出一种基于遗传博弈的部分任务卸载算法(Genetic Game Theory-Partial Task Offloading, GGT-PTO)。将云机器人任务完成时间与系统能耗作为衡量系统总代价的两个指标,通过设置任务指标权重来模拟实际任务需求偏好,使用两个阈值对任务进行粒化分割。在每一轮对系统总代价的博弈中,使用遗传算法代替常规设置步长的方式,选出能使系统总代价降低量最大化的单个任务,并将该任务此时对应的卸载阈值更新入系统卸载策略集合中。不断重复以上过程,寻找该算法下的纳什平衡状态,确定整个系统的最佳卸载策略集合,进而降低整个系统的总代价。仿真结果表明,相比于常见二进制卸载算法和基于传统博弈的部分卸载算法,该算法在任务完成时间和能耗上均有明显降低,可使云边端协同服务的整体性能显著提升。

How to make full use of the computing resources of the central cloud, edge cloud and local devices to reduce the total cost of the system as much as possible is one of the main problems that need to be solved in the field of cloud robot task unloading. To solve this problem, a genetic game-based partial task offloading algorithm (GGT-PTO) is proposed. The cloud robot task completion time and system energy consumption are taken as two indicators to measure the total cost of the system. The weight of the task indicator is set to simulate the actual task demand preference, and two thresholds are used to granulate the task. In the game of the total cost of the system in each round, the genetic algorithm is used instead of the conventional method of setting the step size to select a single task that can maximize the total cost reduction of the system, and the corresponding unloading threshold of the task at this time is updated into the system unloading strategy set. Repeat the above process, find the Nash equilibrium state under the algorithm, determine the best unloading strategy set of the whole system, and then reduce the total cost of the whole system. The simulation results show that compared with other binary offloading algorithms and partial offloading algorithms based on traditional games, the algorithm significantly reduces the task completion time and energy consumption, and can significantly improve the overall performance of cloud-edge collaborative services.